Ryan Tannehill

"Etrius24 wrote:

Ugh

Of course the EPA was low for Tannehill in 2018... Gase was running the offense... This has been discussed... Gase went to the Jets and ****ed their offense up also... 3 yard passing plays... Running the same ineffective play over and over again... Predictable play calling.

I am disappointed Guy.... This crap is weak... It reeks of desperation

The point was more about how his 2019 figure was a mere 0.65 standard deviations above league average, despite that his passer rating was 2.29.

Among the QB seasons since 2017 that have been better in terms of EPA per play:

Case Keenum in 2017
Mitchell Trubisky in 2018
Jameis Winston in 2018
Dak Prescott in 2017
Dak Prescott in 2019
Alex Smith in 2017

And by the way, I've moved on from trying to convince anybody of anything here. Only 2020 will adjudicate this issue. Nobody is budging at this point. The stuff above is merely informational.

"The Guy wrote:

The take-home message is this -- his adjusted passer rating was 2.29 standard deviations above the league average from 2017 to 2019, but his EPA per play again only 0.65.

And? I'll take the QB that led the league in passer rating, YPA, and CPOE.

"The Guy wrote:

The point was more about how his 2019 figure was a mere 0.65 standard deviations above league average, despite that his passer rating was 2.29.

Among the QB seasons since 2017 that have been better in terms of EPA per play:

Case Keenum in 2017
Mitchell Trubisky in 2018
Jameis Winston in 2018
Dak Prescott in 2017
Dak Prescott in 2019
Alex Smith in 2017

And by the way, I've moved on from trying to convince anybody of anything here. Only 2020 will adjudicate this issue. Nobody is budging at this point. The stuff above is merely informational.

Looks to me like a terribly flawed stat when used in isolation.

"The Guy wrote:

And by the way, I've moved on from trying to convince anybody of anything here. Only 2020 will adjudicate this issue.

Better hope that coronavirus doesn't cancel the season! Otherwise this debate will continue unabated into 2021 LOL.

OK.. so it's probably useful if you say beforehand what your conclusions will be if let's say Tannehill starts a minimum of 6 games and:
1) makes the top 10 but not top 6 in passer rating
2) makes the top 6 in passer rating

If he fails to make the top 10 it's obvious: 2019 was truly an aberration like you occasionally see (e.g. Dalton in 2015 where he came in 2nd in passer rating). But what will you argue if #1 or #2 happens?

We should have this on record beforehand. Remember, I'll just go by statistical significance tests whatever they show.

"FinFaninBuffalo wrote:

Looks to me like a terribly flawed stat when used in isolation.

The issue is that in passer rating and YPA, all yards "are created equal" so to speak, whereas for EPA the following is the case:

The value of a football play has traditionally been measured in yards gained. Unfortunately, yards is a flawed measure because not all yards are equal. For example, a 4-yard gain on 3rd down and 3 is much more valuable than a 4-yard gain on 3rd and 8. Any measure of success must consider the down and distance situation.

Field position is also an important consideration. Yards gained near the goal line are tougher to come by and are more valuable than yards gained at midfield. Yards lost near one’s own goal line can be more costly as well.

We can measure the values of situations and, by extension, the outcomes of plays by establishing an equivalence in terms of points. To do this we can start by looking back through recent NFL history at the ‘next points scored’ for all plays. For example, if we look at all 1st and 10s from an offense’ own 20-yard line, the team on offense will score next slightly more often than its opponent. If we add up all the ‘next points’ scored for and against the offense’s team, whether on the current drive or subsequent drives, we can estimate the net point advantage an offense can expect for any football situation. For a 1st and 10 at an offense’s own 20, it’s +0.4 net points, and at the opponent’s 20, it’s +4.0 net points. These net point values are called Expected Points (EP), and every down-distance-field position situation has a corresponding EP value.

Suppose the offense has a 1st and 10 at midfield. This situation is worth +2.0 EP. A 5-yard gain would set up a 2nd and 5 from the 45, which corresponds to a +2.1 EP. Therefore, that 5-yard gain in that particular situation represents a +0.1 gain in EP. This gain is called Expected Points Added (EPA). Likewise, a 5-yard loss on 1st down at midfield would create a 2nd and 15 from the offense’s own 45. That situation is worth +1.2 EP, representing a net difference of -0.8 EPA.

https://www.advancedfootballanalytics.com/index.php/home/stats/stats-explained/expected-points-and-epa-explained

"The Guy wrote:

The issue is that in passer rating and YPA, all yards "are created equal" so to speak, whereas for EPA the following is the case:

https://www.advancedfootballanalytics.com/index.php/home/stats/stats-explained/expected-points-and-epa-explained

Any stat that puts any season by Winston, Tribusky, and Keenum ahead of what Tannehill did last season is fatally flawed.

"The Guy wrote:

The issue is that in passer rating and YPA, all yards "are created equal" so to speak, whereas for EPA the following is the case:

https://www.advancedfootballanalytics.com/index.php/home/stats/stats-explained/expected-points-and-epa-explained

EPA is a solid concept, but are you using actual EPA per play or is this ESPN's "clutch-weighted" EPA? Because that could make a huge difference since we don't know what the "clutch" formula is doing.

I ask because you can definitely get EPA data for a QB but so far I've only found it in play-by-play data. I don't know of a site that just tells you EPA for a full QB season (maybe you know of one?). The ones I've seen use ESPN's clutch-weighted EPA, and if that's the case there's reason to suspect ESPN might be messing with things in a way you wouldn't want.

"cbrad wrote:

EPA is a solid concept, but are you using actual EPA per play or is this ESPN's "clutch-weighted" EPA? Because that could make a huge difference since we don't know what the "clutch" formula is doing.

I ask because you can definitely get EPA data for a QB but so far I've only found it in play-by-play data. I don't know of a site that just tells you EPA for a full QB season (maybe you know of one?). The ones I've seen use ESPN's clutch-weighted EPA, and if that's the case there's reason to suspect ESPN might be messing with things in a way you wouldn't want.

I was wondering where he is getting his data from as well....

"cbrad wrote:

Better hope that coronavirus doesn't cancel the season! Otherwise this debate will continue unabated into 2021 LOL.

OK.. so it's probably useful if you say beforehand what your conclusions will be if let's say Tannehill starts a minimum of 6 games and:
1) makes the top 10 but not top 6 in passer rating
2) makes the top 6 in passer rating

If he fails to make the top 10 it's obvious: 2019 was truly an aberration like you occasionally see (e.g. Dalton in 2015 where he came in 2nd in passer rating). But what will you argue if #1 or #2 happens?

We should have this on record beforehand. Remember, I'll just go by statistical significance tests whatever they show.

If we're doing this with passer rating exclusively, my prediction is that his 2020 season will be non-significantly different from his pre-2019 performance, statistically speaking.

"The Guy wrote:

If we're doing this with passer rating exclusively, my prediction is that his 2020 season will be non-significantly different from his pre-2019 performance, statistically speaking.

OK, but what will you conclude if it IS statistically significant?

The reason I ask is because statistical significance alone won't actually answer the question of how much of the improvement was due to the surroundings vs. Tannehill himself (the crux of the debate here). I'm just wondering what you'd argue if it is once again statistically significant? Will you still say Tannehill is mostly an average QB but in perfect surroundings? Or will you say it's actually more likely Tannehill is above average but was held back in Miami? The stats alone don't answer that question.

Personally, I only use words like "average" or "elite" in this debate to represent actual performance, so when I use stats I sidestep that issue (precisely because I don't see how current stats can answer it).

"FinFaninBuffalo wrote:

Any stat that puts any season by Winston, Tribusky, and Keenum ahead of what Tannehill did last season is fatally flawed.

There may be something to it, however, because when we subtract the z-score for EPA from the z-score for adjusted (to 2019) passer rating and then find the z-scores of those differences -- again for the sample of 93 QB seasons from 2017 to 2019 -- we find that Tannehill leads the league in that regard, on both of his occasions in the sample, in 2018 and 2019.

He's slightly above a whopping three standard deviations from the mean in both instances (3.04 and 3.01), and there are but seven of the 93 QB seasons during the period from 2017 to 2019 who are at least two standard deviations above the mean.

So whatever is going on that makes Tannehill's EPA so much different from his passer rating, it's now happened in multiple seasons, both a good one and a bad one, and he's far and away beyond the league norm in that regard.

"FinFaninBuffalo wrote:

Any stat that puts any season by Winston, Tribusky, and Keenum ahead of what Tannehill did last season is fatally flawed.

Its flawed because it punishes for failure on first and second down that may have nothing to do with the QB.

It assumes that a better play was available or asks the QB to hold the ball an indefinite amount of time until it becomes available.

What if on both of those plays the only open man was the 4 yard slant or drag? Why are we punishing what is possibly the correct and only read?

As a QB that isnt a bad decision to begin with. You're hoping your WR can beat one defender and pick up the first.

"cbrad wrote:

EPA is a solid concept, but are you using actual EPA per play or is this ESPN's "clutch-weighted" EPA? Because that could make a huge difference since we don't know what the "clutch" formula is doing.

I ask because you can definitely get EPA data for a QB but so far I've only found it in play-by-play data. I don't know of a site that just tells you EPA for a full QB season (maybe you know of one?). The ones I've seen use ESPN's clutch-weighted EPA, and if that's the case there's reason to suspect ESPN might be messing with things in a way you wouldn't want.

It's the clutch-weighted EPA from ESPN. Apparently they down-weight EPA during garbage time, which is probably defined as some level of win probability at and beyond some point in the game (e.g., 10% or less likely to win, in the fourth quarter), and they don't weight it upward or downward at any other time.

So, because of that, no one can argue that Tannehill's EPA is artificially inflated by garbage time performance, because it isn't inflated in the first place.

"cbrad wrote:

OK, but what will you conclude if it IS statistically significant?

The reason I ask is because statistical significance alone won't actually answer the question of how much of the improvement was due to the surroundings vs. Tannehill himself (the crux of the debate here). I'm just wondering what you'd argue if it is once again statistically significant? Will you still say Tannehill is mostly an average QB but in perfect surroundings? Or will you say it's actually more likely Tannehill is above average but was held back in Miami? The stats alone don't answer that question.

Personally, I only use words like "average" or "elite" in this debate to represent actual performance, so when I use stats I sidestep that issue (precisely because I don't see how current stats can answer it).

I think I would need to see some teasing apart of Tannehill from Henry's performance, amidst a similar season in terms of Tannehill's performance, to regard him as having greater individual ability than I previously thought.

If the correlation between Tannehill's passer rating and Henry's yards per rush, game-by-game, is again far and away the highest in the league, and the team's use of Henry again permits Tannehill to throw so few passes, while he performs significantly more poorly when he does throw more passes, well then it'll be tough to conclude that he has greater individual ability than I'd thought.

What I'd like to see is what happens if and when their performances do become independent of each other.

Guy

Stop lying.

2020 will only adjudicate this issue for you if he fails... As has been pointed out by all of the other posters here. You have been on this crusade for years getting banned from other sites because of it. You were hell bent on proving these things before the 2019 season... Tannehill went out there and killed it and you just tweaked your arguments and proceeded forward with your cause.

All evidence points to you merely changing your argument and finding other stats to try to prove your point if Tannehill has a successful 2020 season.

"AGuyNamedAlex wrote:

Its flawed because it punishes for failure on first and second down that may have nothing to do with the QB.

No it actually accounts for that, because passes on third down, after failures on first and second down, are associated with a relatively low number of expected points. So when the QB does what's expected on third down, after failures on first and second, he isn't penalized. The EPA in that situation is near zero.

In other words, not much is "expected" of a QB on third down, after there were failures on first and second.

"Etrius24 wrote:

Guy

Stop lying.

2020 will only adjudicate this issue for you if he fails... As has been pointed out by all of the other posters here. You have been on this crusade for years getting banned from other sites because of it. You were hell bent on proving these things before the 2019 season... Tannehill went out there and killed it and you just tweaked your arguments and proceeded forward with your cause.

All evidence points to you merely changing your argument and finding other stats to try to prove your point if Tannehill has a successful 2020 season.

If all that were true then I would be highly unwilling to put anything in that regard in writing right now, when it can easily be referred back to after next season.

"The Guy wrote:

I think I would need to see some teasing apart of Tannehill from Henry's performance, amidst a similar season in terms of Tannehill's performance, to regard him as having greater individual ability than I previously thought.

If the correlation between Tannehill's passer rating and Henry's yards per rush, game-by-game, is again far and away the highest in the league, and the team's use of Henry again permits Tannehill to throw so few passes, while he performs significantly more poorly when he does throw more passes, well then it'll be tough to conclude that he has greater individual ability than I'd thought.

What I'd like to see is what happens if and when their performances do become independent of each other.

Right.. so 2020 won't settle it unless Tannehill performs similar (statistically) to 2012-2018.

That's a problem. It's important to lay out how to test a hypothesis, and you're making it very hard to reject the point of view you have.

"cbrad wrote:

Right.. so 2020 won't settle it unless Tannehill performs similar (statistically) to 2012-2018.

That's a problem. It's important to lay out how to test a hypothesis, and you're making it very hard to reject the point of view you have.

No he could perform similar to how he did in 2019, but in a way that's far less strongly associated with Henry. That would make me revise my appraisal of his ability.

"The Guy wrote:

No he could perform similar to how he did in 2019, but in a way that's far less strongly associated with Henry. That would make me revise my appraisal of his ability.

But there's the problem. Correlation doesn't imply causation, and you're assuming that all the effect of that correlation is due to Henry to test whether Tannehill's passer rating was due to Henry or Tannehill?

Can't do that. The test has to be independent of the hypothesis you're trying to test.

btw.. that IS confirmation bias.

"cbrad wrote:

btw.. that IS confirmation bias.

Told you.....

"cbrad wrote:

But there's the problem. Correlation doesn't imply causation, and you're assuming that all the effect of that correlation is due to Henry to test whether Tannehill's passer rating was due to Henry or Tannehill?

Can't do that. The test has to be independent of the hypothesis you're trying to test.

btw.. that IS confirmation bias.

No what I'm saying is that another strong association between their performances provides inconclusive data regarding any change in the appraisal of Tannehill's individual ability. I'm not saying it provides another year of conclusive data consistent with 2012-2018.

"The Guy wrote:

No what I'm saying is that another strong association between their performances provides inconclusive data regarding any change in the appraisal of Tannehill's individual ability. I'm not saying it provides another year of conclusive data consistent with 2012-2018.

I didn't say that this hypothetical scenario would mean 2012-2018 was similar. I'm just saying it IS confirmation bias. You're not considering all kinds of different hypotheses and then filtering them out based on evidence. You're starting with one specific hypothesis (the one you prefer) that's so hard to test that it's really unlikely you'll be "proven wrong" in 2020.

So saying 2020 will determine this is simply wrong. There's a decent chance it's (as you say) "inconclusive".

"cbrad wrote:

I didn't say that this hypothetical scenario would mean 2012-2018 was similar. I'm just saying it IS confirmation bias. You're not considering all kinds of different hypotheses and then filtering them out based on evidence. You're starting with one specific hypothesis (the one you prefer) that's so hard to test that it's really unlikely you'll be "proven wrong" in 2020.

So saying 2020 will determine this is simply wrong.

Hold on -- what we have here is just one season that stands in contrast with his others, and during that season he 1) had by far the strongest correlation in the league between his performance and that of his running back, 2) threw the second-fewest number of passes per game in the league, which was a much lower number than he threw from 2012 to 2018, and 3) performed far more poorly when the situation dictated that he throw more passes.

I don't see how such a season can permit a conclusive change in previous appraisals of his individual ability. And then if he has just another season of that type, I don't see how that one can do it either.

"The Guy wrote:

I don't see how such a season can permit a conclusive change in previous appraisals of his individual ability. And then if he has just another season of that type, I don't see how that one can do it either.

Exactly!

That's all I wanted you to admit. You started this off (many times actually) by saying 2020 will determine this. All I'm pointing out is you've created a hypothesis that's easy to confirm IF Tannehill performs average, but really hard to disprove (not impossible, but really hard).

Point is.. let's stop with this notion that 2020 will determine things. There's a good chance it won't given the hypothesis you've chosen.